WPS4982 Policy Research Working Paper 4982 More on the Energy/Non-Energy Commodity Price Link John Baffes The World Bank Development Prospects Group Global Trends Team June 2009 Policy Research Working Paper 4982 Abstract This paper examines the energy/non-energy commodity contrast, only a few price indices responded strongly to price link, based on a reduced form econometric inflation, although the trend parameter estimate (often model and using annual data from 1960 to 2008. The viewed as a proxy for technological progress) is negative transmission elasticity from energy to the non-energy for agriculture and positive for metals. A key implication index is estimated at 0.28. At a more disaggregated of the pass-through results is that for as long as energy level, the fertilizer index exhibited the largest elasticity prices remain elevated, most non-energy commodity (0.55), followed by precious metals (0.46), food (0.27), prices are expected to be high. metals and minerals (0.25), and raw materials (0.11). By This paper--a product of the Global Trends Team, Development Prospects Group--is part of a larger effort in the department to analyze and understand the key determinants of commodity price movements. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at jbaffes@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team More on the Energy/NonEnergy Commodity Price Link John Baffes THE WORLD BANK 1818 H Street, NW Washington, D.C. 20433 tel: (202) 4581880 fax: (202) 5221151 email: jbaffes@worldbank.org Follow up to WPS 4333 JEL: O13, Q33 KEY WORDS: energy prices, nonenergy prices, transmission elasticities, passthrough I. Introduction By most accounts, the recent commodity boom was the longest and broadest (in terms of commodities involved) of the postWWII period (World Bank 2009). Between 2003 and 2008, nominal energy and metal prices increased by 230%, food and precious met als doubled, while fertilizer prices increased fourfold. Although prices (except precious metals) have declined sharply since their mid2008 peak, as of mid2009 they are still considerably higher than their lows reached in early 2000s. The breadth of the recent boom is generating renewed interest on the nature and degree of commodity price links. This paper focuses on one such link: the price trans mission from energy to nonenergy commodities. Specifically, 11 nonenergy commodi ty price indices are regressed on an energy price index, a measure of inflation, and a time trend (considered as a proxy to technological change) and the corresponding esti mates of transmission elasticities are derived. The paper expands earlier work (Baffes 2007) by first broadening the definition of energy (i.e., use of an energy index which in cludes natural gas and coal in addition to crude oil) and second extending the sample to 2008. The rest of the paper begins with a brief review of the literature on price comove ment. It is then followed by a discussion of the methodology and description of the da ta. The results are discussed next while the last section concludes. II. Literature review and modeling framework Although the literature on the energy/nonenergy commodity price links is relatively thin, the broader subject of price comovement has been examined extensively and in various contexts with the research falling largely within two strands. The first strand examines comovement among prices of the same commodity in different locations with in the market efficiency context, also known as spatial market integration or law of one price (see Fackler and Goodwin (2001) for a literature review). A more important but --2-- less researched subject within that strand has been the comovement between world and domestic commodity prices. This relationship includes the policy dimension by analyz ing whether world price signals have been fully transmitted to domestic markets or such signals have been subjected to policy distortions (see, for example, Baffes and Gardner (2003) and Mundlak and Larson (1992)).1 The second strand of literature examines price comovement (or lack thereof) of different commodities. It goes back to Granger (1986, p. 218) who wrote: "If xt and yt are a pair of prices from a jointly efficient, speculative market, they cannot be cointegrated ... if the two prices were cointegrated, one can be used to help forecast the other and this would contradict the efficient market assumption. Thus, for example, gold and sil ver prices, if generated by an efficient market, cannot move closely together in the long run." Granger's assertion led to research in commodity markets (e.g., MacDonald and Taylor 1988) and other markets as well, notably exchange rates (see, among others, Bail lie and Bollerslev (1989) and Hakkio and Rush (1989)). This research was later ques tioned on several grounds including the fact that comovement reflects response to common fundamentals rather than market inefficiencies. See, for example, Agbeyebbe (1992), Baffes (1993), Dwyer and Wallace (1992), and Sephton and Larsen (1991). A similarly controversial subject has been the socalled `excess comovement' hy pothesis first discussed by Pindyck and Rotemberg (1990) who, after analyzing price movements of seven seemingly unrelated commodities (cocoa, copper, cotton, crude oil, gold, lumber, and wheat), they concluded that these prices comoved in excess of what the macroeconomic variables could explain. A number of likely explanations were giv en for such comovement including incomplete model, endogeneity of the macroeco nomic variables, rejection of normality assumption, and bubbles or market psychology.2 Subsequent research, however, challenged the excess comovement hypothesis on data and methodological grounds (see Ai, Chatrath, and Song (2006), Cashin, McDermott, --3-- and Scott (1999), Deb, Trivedi, and Varangis (1996), and LeyBourne, Lloyd, and Reed (1994)).3 The present paper examines the price transmission from energy to nonenergy commodities. Specifically, it estimates transmission elasticities for 11 nonenergy price indices based on the following specification: log(NON_ENERGYt) = + 1log(ENERGYt) + 2log(MUVt) + 3TIME + t, where NON_ENERGYt denotes the various nonenergy US dollarbased commodity price indices at time t, ENERGYt denotes the energy price index, MUVt denotes the deflator, TIME is time trend, and t denotes the error term, the properties of which will be subject to empirical investigation; , 1, 2, and 3 denote parameters to be estimated. Because the model is expressed in logarithms, the parameter estimates can be interpreted as elasticities. Although the signs and magnitudes of the coefficients are not dictated by eco nomic theory, 1 and 2 are expected to be positive because energy as well as other goods and services (as reflected by the measure of inflation) constitute key inputs to the production process of all commodities.4 (See Baffes (2007), FAO (2002), and World Bank (2009) on the various transmission channels between energy and nonenergy markets; Kilian (2008) on the effects of energy shocks to the broader economy; and Schmidhuber (2006) on the impact of biofuels on agricultural markets.) On the other hand, 3 is ex pected to be negative, at least for agriculture--consistent with the long term impact of technological progress on production costs as well as the low income elasticity of most food commodities, especially cereals. As an alternative to the above specification, one could deflate both indices by di viding NON_ENERGYt and ENERGYt with MUVt, effectively restricting the sum of the energy index and inflation coefficients to unity (i.e., 1 + 2 = 1). Estimating nominal in dices, however, relaxes the homogeneity restriction so that a direct estimate of the im --4-- pact of inflation can be obtained (Houthakker 1975). Annual data for 11 nonenergy indices covering the period from 1960 to 2008 are used in the analysis, giving a total of 49 observations. The Manufacture Unit Value (MUV) is used as an inflation proxy. The MUV, often considered as a developed coun try deflator indicator, represents the unit value index in dollar terms of manufactures exported from industrial countries weighted proportionally to the countries' exports to the developing countries. Data on all indices are collected and reported by the World Bank (World Bank, various issues) and are defined as follows. Energy [world tradebased weights in square brackets] consists of crude oil [84.6%], natural gas [10.8%], and coal [4.6%]. Nonenergy consists of metals [31.6%], fertilizers [3.6%], and agriculture [64.8%]. Agriculture con sists of beverages [8.4%], raw materials [16.5%], and food [39.9%] while Food consists of cereals [11.3%], edible oils [16.3%], and other food [12.3%]. The commoditycomposition of the subindices is as follows: Beverages: cocoa, coffee (arabica and robusta), tea; Ce reals: maize, rice, sorghum, wheat; Edible oils: coconut oil, groundnut oil, palm oil, soybean meal, soybean oil, soybeans; Other food: bananas, meat (beef and chicken), oranges, shrimp, sugar; Raw materials: cotton, rubber, timber (logs and sawnwood); Fertilizers: DAP, phosphate rock, potassium chloride, TSP, urea; Metals: aluminum, copper, iron ore, lead, nickel, tin, zinc. Finally, a Precious metal index was included in the analysis--not part of the nonenergy index--consisting of gold [89.7%] and silver [10.3%]. The use of low frequency data was motivated by the desire to circumvent the noise typically embedded in high frequency data and also compare the effect of energy prices on the prices of all primary commodity groups. Because most agricultural com modities are subject to crop cycles, annual frequency is, perhaps, more relevant. For ex ample, the decision of how much land to allocate to each commodity and how much in --5-- puts to use is taken once a year, typically prior to planting. On the other hand, although a higher frequency would add more observations, the high variability characterizing commodity prices implies that even annual observations contain a large amount of in formation (Campos and Ericsson 1999). III. Results First, I examined the stationarity properties of all the series in loglevel form using the augmented DickeyFuller (ADF) testing procedure (these statistics are not reported here). With the exception of the cereals index (ADF = 1.76, significant at 10%), the re maining AFD statistics ranged from 0.15 (metals) to 1.53 (other food). Appending a time trend to the ADF regressions did not alter the statistics in any significant way thus confirming the existence of a unit root. In other words, nominal commodity price indic es do not exhibit a strong meanreverting process nor they move around a linear trend; instead they are characterized by a long memory process. Differencing the series once, however, induced stationarity in all cases, implying that validating the model would require examination of the stationarity properties of its error term, in addition to con ventional indicators such as R2s and tratios. Recall that nonstationarity of price series is not a surprising result considering the long period along with the low frequency of the data as well as the fact that the series were expressed in nominal terms. Results from OLS regressions are presented in table 1. Specifically, the first four columns report parameter estimates of the constant term, energy, inflation, and the time trend, followed by the adjustedR2 and the ADF statistic. The estimates--reflected in the sign of the energy price coefficient as well as the conventional and stationarity statis tics--indicate that energy prices and to a lesser extent inflation and technological change explain a considerable part of commodity price variability (the adjusted R2 of all regressions averaged 0.85). Moreover, the ADF statistics were in all but one case (edible oils) far below 3.00 (they averaged 3.77), fully consistent with a stationary error term. --6-- The parameter estimate of the nonenergy index (top row of table 1) is 0.28, im plying that a 10% increase in energy prices is associated with a 2.8% increase in non energy commodity prices, in the long run. Three earlier studies--Gilbert (1989), Bo rensztein and Reinhart (1994), and Baffes (2007)--reported elasticities of 0.12, 0.11, and 0.16, respectively (see table 2). When the sample of the current analysis is adjusted to match the samples of these studies, the passthrough coefficient becomes remarkably similar (0.13 and 0.12, and 0.18, respectively). The transmission elasticity of the nonenergy index, however, masks some varia tions. The highest elasticity among the subindices was in fertilizer, estimated at 0.55, not surprisingly since nitrogenbased fertilizers are made directly from natural gas. Fer tilizer and energy price increases during the recent boom were in line with the ones ex perienced during the first oil shock: from 1973 to 1974 phosphate rock and urea prices increased fourfold and threefold in nominal terms, remarkably similar to the crude oil nominal price increase during that period, from $2.81/barrel to $10.97/barrel. The transmission elasticity for agriculture, estimated at 0.27, reflects an average of wide ranging parameter estimates: beverages (0.38), food (0.27) and raw materials (0.11). The elasticity estimates of the food price index components, however, fall within a very narrow range: cereals (0.28), edible oils (0.29), and other food (0.22), all signifi cantly different from zero at the 1% level.5 The metals elasticity was estimated at 0.25; yet its components presented considerable variation (individual commodity elasticity estimates are not reported here). Note that highly diverse estimates in the passthrough elasticities of metal prices were the key findings of Baffes (1997, 2007) and Chaudhri (2001). The precious metals elasticity was estimated at 0.46--the second largest among the 11 subindices studied here. Its large value reflects the association of high energy prices with inflationary pressures, slower economic growth, and resource scarcity, all of --7-- which prompt households and investors to view precious metals (especially gold) as safe investment alternatives, therefore increasing their demand and hence their prices. Not surprisingly, the two postBretton Woods peaks of gold prices, $750/toz in 1980 and $687/toz in 2008, correspond to the two crude oil price peaks, $45/barrel and $76/barrel, respectively (all prices are expressed in 2000 real US dollars). Three important conclusions emerge from the analysis. First, the results show that the prices of most commodities, and especially those of fertilizers and precious metals, respond firmly to energy prices. Furthermore, such response appears to be strengthening in periods of high commodity prices as confirmed by the fact that the values of the estimated elasticities increase considerably when the recent boom is in cluded in the analysis (in some cases the elasticities double; see difference between the last two columns of table 2). The implication is that, for as long as energy prices remain elevated, analysis of nonenergy commodity markets requires understanding of the energy markets as well. On the other hand, the nonenergy elasticity is insensitive to the model structure and frequency of the data as can be inferred by its remarkable similari ty with the earlier studies (when adjusted for sample size). Second, while the transmission elasticities were broadly similar, this was not the case with the inflation coefficient, 2, the estimates of which varied considerably in terms of sign, magnitude, and level of significance. It was positive and significantly dif ferent from zero only for precious metals and agriculture (including two of its sub indices) while it was effectively zero for metals and fertilizers. All this implies that the relationship between inflation (at least as measured by the MUV) and nominal com modity prices is much more complex and, perhaps, changing over time. This may not be surprising if one considers that during 197280 (a period which includes both oil shocks) the MUV increased by 45% while during 200008, it increased by half as much. The nominal nonenergy price index increase during these two 8year periods was iden --8-- tical at 170%. Third, the trend parameter estimates, 3, are spread over an even wider range compared to energy passthrough and inflation parameter estimates. The nonenergy price index, for example, shows no trend at all. Yet, the metal price index exhibited an almost 2% positive annual trend while the agriculture index showed a 1% negative an nual trend. Furthermore, the trend parameter estimates of the agriculture subindices vary considerably both in magnitude and level of significance, from 0.08 (tvalue = 0.19) for raw materials to 3.12 (tvalue = 5.22) for beverages, a result consistent with Deaton's (1999, p. 27) observation that what commodity prices lack in trend, they make up in va riability. On the other hand, the large variation among the trend parameter estimates implies that the validity of the PrebischSinger hypothesis, often discussed in the con text of the secular decline of primary commodity prices, may require some rethinking (see Spraos 1980, among others). IV. Concluding remarks Based on 19602008 annual data and a simple econometric model, I estimated longrun energy transmission elasticities for 11 nonenergy commodity groups. The elasticity for the nonenergy index was estimated at 0.28. At a disaggregated level, the fertilizer index exhibited the largest elasticity, followed by precious metals, food, and metals & miner als. The key implication of these findings is that, for as long as energy prices remain elevated, most nonenergy commodity prices are expected to remain high and thus any analysis of nonenergy commodity markets cannot be undertaken in isolation to devel opments in energy markets. On the methodological side, the fact that the estimates of the current study (which included the recent commodity boom) are larger than earlier ones (not account ing for the recent boom) implies that, perhaps, timevarying parameter or switching --9-- regime models may be more appropriate in analyzing the energy/nonenergy commodi ty price links. Such models could shed more light on the relationship between inflation and commodity prices; they may also enhance our understanding on the (well researched but not yet settled) subject of the secular decline of primary commodity prices. Another avenue to pursue is, naturally, the use of higher frequency data within an errorcorrection framework. -- 10 -- ENDNOTES 1 The thinly researched subject of world/domestic price comovement reflects, perhaps, the un availability of data. That, however, changed recently for agricultural products following the re search project led by Kym Anderson whose methodology (Anderson et al. 2008) resulted in a consistent global database which includes prices received by farmers and paid by consumers in 75 countries (www.worldbank.org/agdistortions). 2 The rejection of the efficient market hypothesis in the presence of comovement argued by Granger (1986) corresponds to Pindyck and Rotemberg's (1990) `bubbles' or `market psycholo gy' explanation for excess comovement--provided that prices used in Granger's sense have been adjusted accordingly by the fundamentals. 3 The early methodological setbacks of the literature of price comovement among different commodities may explain why the subject has not been researched more thoroughly. 4 The exogeneity of energy prices assumed here reflects the large size of energy markets com pared to the size of other commodity markets as well as the fact that energy is a key input in the production process of most commodities. Hence, 1 is interpreted as transmission elasticity ra ther than just a cointegration parameter, similar to the cases of domestic/world price links where domestic prices are typically assumed to be a function of world prices. 5 Commodityspecific regressions show that the high elasticity estimate of beverages is driven by cocoa (1 = 0.52, tvalue = 5.35), a result with no obvious explanation. On the other hand, the similar elasticity size among food subindices extends to most individual commodities especial ly for the components of grain and edible oil indices. -- 11 -- TABLE 1: PARAMETER ESTIMATES INDEX 1 2 100*3 AdjR2 ADF NonEnergy 3.03@ 0.28@ 0.12 0.01 0.90 3.35** (6.54) (5.24) (0.68) (0.02) Metals 3.77@ 0.25@ 0.17 1.93@ 0.82 3.30** (4.80) (3.14) (0.60) (2.31) Fertilizers 3.58@ 0.55@ 0.30 0.39 0.81 3.97*** (4.12) (4.79) (0.95) (0.48) Agriculture 2.51@ 0.26@ 0.33@ 0.99@ 0.90 3.81*** (6.90) (5.54) (2.43) (2.73) Beverages 1.83@ 0.38@ 0.55@ 3.12@ 0.76 4.95*** (3.10) (4.87) (2.63) (5.22) Raw materials 1.85@ 0.11@ 0.51@ 0.08 0.91 3.15** (4.16) (2.15) (3.15) (0.19) Food 2.91@ 0.27@ 0.21 0.71 0.85 3.85*** (7.11) (4.93) (1.39) (1.80) Cereals 3.13@ 0.28@ 0.17 0.87 0.78 3.83*** (5.94) (4.23) (0.89) (1.76) Edible oils 3.33@ 0.29@ 0.12 0.80 0.80 2.82* (6.16) (4.51) (0.58) (1.50) Other food 1.86@ 0.22@ 0.45@ 0.42 0.89 3.60*** (6.28) (3.81) (4.44) (1.18) Precious metals 1.40@ 0.46@ 1.05@ 1.75@ 0.98 3.91*** (3.58) (9.40) (7.61) (3.68) Notes: The @ sign denotes parameter estimate significant at the 5% level while the numbers in parentheses are absolute tvalues (the corresponding variances have been estimated using White's method for heteroskedasticityconsistent standard errors.) ADF denote the MacKinnon onesided pvalues based on the Augmented DickeyFuller equation (Dickey and Fuller 1979). One (*), two (**), and three (***) asterisks indicate rejection of the existence of one unit root at the 10%, 5%, and 1% levels of significance (the respective tstatistics are 2.60, 2.93, and 3.58). The lag length of the ADF equations was determined by minimizing the Schwarzloss function. Source: Author's estimates. -- 12 -- TABLE 2: COMPARING LONGRUN TRANSMISSION ELASTICITIES Holtham (1988) Gilbert (1989) Borensztein & Baffes (2007) This Study 1967:S11984:S2 1965:Q11986:Q2 Reinhart (1994) 19602005 19602008 1970:Q11992:Q3 Nonenergy -- 0.12 0.11 0.16 0.28 Food -- 0.25 -- 0.18 0.27 Raw materials 0.08 -- -- 0.04 0.11 Metals 0.17 0.11 -- 0.11 0.25 Notes: Holtham uses semiannual data, Gilbert and Borensztein & Reinhart quarterly, and Baffes along with the present study annual. Gilbert's elasticities denote averages based of four specifi cations. Holtham's raw materials elasticity is an average of two elasticities based on two sets of weights. `--` indicates that the estimate is not available. Source: Holtham (1988), Gilbert (1989), Borensztein and Reinhart (1994), Baffes (2007), and au thor's estimates. -- 13 -- REFERENCES Agbeyebbe, Terence D. (1992). "Common Stochastic Trends: Evidence from the London Metal Exchange." Bulletin of Economic Research, vol. 44, pp. 141151. Ai, Chunrong, Arjun Chatrath, and Frank Song (2006). "On the Comovement of Commod ity Prices." American Journal of Agricultural Economics, vol. 88, pp. 574­588. Anderson, Kym, Marianne Kurzweil, William Martin, Damiano Sandri, and Ernesto Va lenzuela (2008). "Measuring Distortions to Agricultural Incentives, Revisited." World Trade Review, vol. 7, pp. 675704. Baffes, John (2007). "Oil Spills on other Commodities." Resources Policy, vol. 32, pp. 126134. Baffes, John (1997). "Explaining Stationary Variables with Nonstationary Regressors." Ap plied Economics Letters, vol. 4, pp. 6975. Baffes, John (1993). "Does Comovement Among Exchange Rates Imply Market Inefficien cy?" Economics Letters, vol. 44, pp. 273280. Baffes, John and Bruce Gardner (2003). "The Transmission of World Commodity Prices to Domestic Markets under Policy Reforms in Developing Countries." Journal of Policy Reform, vol. 6, pp. 159180. Baillie, Richard T. and Tim Bollerslev (1989). "Common Stochastic Trends in a System of Exchange Rates." Journal of Finance, vol. 44, pp. 167181. Borensztein, Eduardo and Carmen M. Reinhart (1984). "The Macroeconomic Determinants of Commodity Prices." IMF Staff Papers, vol. 41, pp. 236261. Campos, Julia and Neil R. Ericsson (1999). "Constructive Data Mining: Modeling Consum ers' Expenditures in Venezuela." Econometrics Journal, vol. 2, pp. 226240. Cashin, Paul, C. John McDermott, and Alasdair Scott (1999). The Myth of Comoving Commodity Prices. IMF Working Paper 169. Chaudhri, Kausik (2001). "LongRun Prices of Primary Commodities and Oil Prices." Ap plied Economics, vol. 33, pp. 531538. Deaton, Angus (1999). "Commodity Prices and Growth in Africa." Journal of Economic Pers pectives, vol. 13, pp. 2340. Deb, Partha, Pravin K. Trivedi, and Panayotis Varangis (1996). The Excess Comovement of Commodity Prices Reconsidered. Journal of Applied Econometrics, vol. 11, pp. 275­91. Dickey, David and Wayne A. Fuller (1979). "Distribution of the Estimators for Time Series Regressions with Unit Roots." Journal of the American Statistical Association, vol. 74, pp. 427 431. Dwyer, Gerald P. and Myles S. Wallace (1992). "Cointegration and Market Efficiency." Journal of International Money and Finance, vol. 11, pp. 318327. Fackler, Paul L. and Barry K. Goodwin (2001). "Spatial Price Analysis." In Handbook of Agricultural Economics, vol. I, ed. B. Gardner and G. Rausser. New York: Elsevier Science. -- 14 -- FAO (2002). Commodity Market Review 200102. Rome: Food and Agriculture Organization of the United Nations. Gilbert, Christopher L. (1989). "The Impact of Exchange Rates and Developing Country Debt on Commodity Prices." Economic Journal, vol. 99, pp. 773783. Granger, Clive W. J. (1986). "Developments in the Study of Cointegrated Economic Va riables." Oxford Bulletin of Economics and Statistics, vol. 48, pp. 213228. Hakkio, Graig S. and Mark Rush (1989). "Market Efficiency and Cointegration: An Appli cation to the Sterling and Deutschmark Exchange Markets." Journal of International Mon ey and Finance, vol. 8, pp. 7588. Holtham, Gerald H. (1988). "Modeling Commodity Prices in a World Macroeconomic Mod el." In International Commodity Market Models and Policy Analysis, ed. Orhan Guvenen. Bos ton: Kluwer Academic Publishers. Houthakker, Hendrik S. (1975). "Comments and Discussion on `The 197275 Commodity Boom' by Richard N. Cooper and Robert Z. Lawrence." Brookings Papers on Economic Ac tivity, vol. 3, pp. 718720. Kilian, Lutz (2008). "The Economic Effects of Energy Price Shocks." Journal of Economic Lite rature, vol. 46, pp. 871909. LeyBourne, Stephen J., Tim A. Lloyd, and Geoffrey V. Reed (1994). "The Excess Com ovement of Commodity Prices Revisited." World Development, vol. 22, pp. 17471758. MacDonald, Ronald and Mark Taylor (1988). "Metals Prices, Efficiency and Cointegration: Some Evidence from the London Metal Exchange." Bulletin of Economic Research, vol. 40, pp. 235239. Mundlak, Yair and Donald L. Larson (1992). "On the Transmission of World Agricultural Prices." The World Bank Economic Review, vol. 6, pp. 399­422. Pindyck, Robert S. and Julio J. Rotemberg (1990). "The Excess Comovement of Commodi ty Prices." Economic Journal, vol. 100, pp. 1173­1189. Schmidhuber, Josef (2006). "Impact of an Increased Biomass Use on Agricultural Markets, Prices and Food Security: A LongerTerm Perspective." International Symposium of Notre Europe, November 2729, Paris. Sephton, Peter S. and Hans K. Larsen (1991). Tests of Exchange Market Efficiency: Fragile Evidence from Cointegration Tests." Journal of International Money and Finance, vol. 10, pp. 561570. Spraos, John (1980). "The Statistical Debate on the Net Barter Terms of Trade between Pri mary Commodities and Manufactures." Economic Journal, vol. 90, pp. 107128. World Bank (2009). Global Economic Prospects: Commodities at the Crossroads. Washington D.C. World Bank (various issues). Commodity price data. Development Prospects Group. Wash ington D.C. -- 15 --